Deep Learning Based Detection Tool for Impacted Mandibular Third Molar Teeth

Area of Science:

• Dentistry
• Medical Imaging
• Artificial Intelligence

Background:

• Impacted third molars are prevalent, leading to complications like decay, root resorption, and pain.
• Accurate detection on panoramic radiographs is crucial for timely intervention and patient care.

Purpose of the Study:

• To develop and evaluate a computer-assisted detection system for impacted third molars using deep convolutional neural networks.
• To compare the performance of different deep learning architectures (Faster R-CNN, YOLOv3) on panoramic radiographs.

Main Methods:

• Utilized a dataset of 440 panoramic radiographs from 300 patients.
• Implemented two-stage (Faster R-CNN with ResNet50, AlexNet, VGG16) and one-stage (YOLOv3) deep learning techniques.
• Evaluated detection performance using metrics like mean Average Precision (mAP@0.5), recall, and precision.

Main Results:

• YOLOv3 achieved the highest detection efficacy with a mAP@0.5 of 0.96, recall of 0.93, and precision of 0.88.
• Faster R-CNN with ResNet50 achieved a mAP@0.5 of 0.91, while VGG16 and AlexNet showed slightly lower performances (0.87 and 0.86).
• The YOLOv3 model demonstrated excellent performance specifically for impacted mandibular third molars.

Conclusions:

• The developed one-stage detector, YOLOv3, shows excellent performance for detecting impacted mandibular third molars on panoramic radiographs.
• Deep learning-based diagnostic tools are reliable and robust for clinical decision-making in dentistry.
• The study highlights the potential of AI in improving the accuracy and efficiency of dental diagnostics.